An introduction to CCC Fiji
Added to website: 15 March 2003
Presented to the South Pacific Applied Geoscience Commission (SOPAC), by James Comley (Director of Marine Science).
Coral Cay Conservation is a not-for-profit organisation whose stated aims are to provide resources to help sustain livelihoods and alleviate poverty through the protection, restoration and management of coral reefs and tropical forests. CCC is self-financed through the use of international volunteers. Operating in this manner, we are able to provide opportunities to our host-county counterparts that would otherwise be prohibitively expensive on an economic and resource basis.
The main output from CCC projects is the production of a detailed habitat map that is based on ground-truthed field data collected by our teams of volunteers. Overlaid onto this basemap is data on the conservation value of the marine habitats identified (Edinger and Risk, 2000), information on fish populations, live hard coral cover and overall univariate biodiversity measures. The end product provides a highly pictorial, visually accessible and readily understood management tool. Ultimately, our in-country counterparts then use this tool in partnership with fundamental socio-economic and demographic data as a basis for making informed management decisions. The marine reserves and MPAs that result are based upon a sound scientific footing, at the heart of which lies the GIS and habitat map created from field survey data extrapolated with the aid of an additional data source in the form of remotely sensed data.
The data analysis and management procedure that CCC use is three fold. An unsupervised classification using an ISOCLASS/ ISOSTAT equivalent algorithm is used to identify spatially confined areas in which spectral reflectance values are homogenous and there it can be assumed that habitat variation will be low. By contrast, areas of high spectral variability are likely to have high habitat heterogeneity. Using this information allows survey strategies to be tailored to concentrate survey effort where it is most needed and therefore make better use of the resources we have at our disposal.
The field data we collect represents a multivariate ‘snapshot’ of the benthic community present and contains a semi quantitative abundance of some 150 sessile organisms and substrates. This data is analysed using the Plymouth Routines In Multivariate Ecological Research (PRIMER) software. Firstly, each record is ordinated against at all others based on the abundance of the species observed in a Multi-Dimensional Scaling (MDS) process. A hierarchal cluster plot based on the Bray-Curtis similarity measure is then constructed which identifies natural groupings of site records- where each of these groups is representative of a habitat. These habitats are given a descriptive name using the data together with the geomorphological area of reef in which they were found. Finally, each habitat is given a quantitative description from the data and other variables such as fish assemblages and conservation values are associated with them. The field data we collect is spatially defined using GPS, allowing a list of GPS locations of examples of each habitat is produced.
Image pre-processing steps we take are the georectification of the imagery, the development of land and deep water masks and, in the case of satellite imagery where there are an appropriate number of spectral bands, atmospheric correction using the Msixs radiometric transfer model and the construction of a water column correction procedure following that proposed by Lyzenga.
The georectification of the image is done using the polynomial, rubber sheet, nearest neighbour resampling algorithm. In the case of multispectral satellite imagery, land and deep-water masks are defined by creating a band-ratioing model relying on the disparity in attenuation of light of different wavelengths in the water column. This method is not possible for tri-spectra aerial imagery. Instead, Areas Of Interest (AOIs) are digitised and manual digital number recoding is done.
The water column correction technique we employ is based upon the physical properties of light of differing wavelengths passing through the water column. Sites of bare substrate of equal reflectance properties, but at different depths are found. Viewed on the remotely sensed image, the disparity in spectral signature between these sites can be attributed solely to the presence of the overlying water column. If enough replicate sites are used, then the ratio of absorption of red versus green versus blue light represents the effect of the absorption of the water column. In this way, a model is constructed that recodes the digital numbers of pixels to remove the spectral contamination of the overlying water.
The next stage we perform following pre-processing is the development of spectral signature sets. The spatially located sites characteristic of each habitat are identified on the image. These ‘seed pixels’ are then used to ‘grow’ AOIs based on user defined differences measured by statistical variability around that of the seed pixel. The resulting spectral signatures of each habitat are examined using the image histograms and the signature set is evaluated and refined in a user-defined iterative process.
Using these signature sets, supervised classification is then performed on the image. The two classification models we use most commonly are the maximum likelihood classifier and the minimum distance with null class removal algorithm. The resulting thematic map then undergoes a process of contextual editing using defined decision rules to recode any obviously misclassified pixels. Finally, a 3x3 median filter is used in the post-classification smoothing of the image. Whilst this reduces may reduce the fine accuracy of the end product, it reduces the pixelated appearance of the image, making it a more user-friendly end product.
The Fiji Reef Conservation Project is a three-year program that has been running since April of 2002 following a successful pilot phase program. Approaching the end of year one, we are currently working towards the production of habitat maps of areas surveyed thus far. The resulting report will contain geographical defined areas of high importance to the restoration and continued management of the Mamanuca Islands.
We are currently engaged in a discussion with other stakeholders throughout Fiji to assess the feasibility of Coral Cay extending its work outside of the present study area.
James Comley
Senior Field Scientist
Coral Cay Conservation
To find out more about SOPAC visit http://www.sopac.org.fj/
